R/convenience.R
In mathesong/pwrcontour: Power Contours And Tables For Different Hypothetical Effect Sizes
Defines functions
pwr.prop.test.table
pwr.prop.test.contour
pwr.r.test.table
pwr.r.test.contour
pwr.t.test.table
pwr.t.test.contour
Documented in
pwr.prop.test.contour
pwr.prop.test.table
pwr.r.test.contour
pwr.r.test.table
pwr.t.test.contour
pwr.t.test.table
#' Plot a Power Contour for a t-test comparison
#'
#' Plot a contour of sample size, effect size and power for t-test comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param esmin Minimum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
#' @param esmax Maximum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the figure are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # A straightforward example
#' pwr.t.test.contour(3, 100, 0.1, 1, sig.level=0.05)
#'
#' # Custom units
#' pwr.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
#' d_unitconversion=5, d_units="mmHg")
#'
#' # Adding things
#' pwr.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
#' d_unitconversion=5, d_units="mmHg") +
#' ggplot2::annotate("text", x=25, y=2,
#' label="We don't want to be here")
pwr.t.test.contour <- function(nmin, nmax, esmin, esmax,
sig.level=0.05,
type = c("two.sample", "one.sample", "paired"),
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
d_unitconversion=1,
d_units="Cohen\'s d",
alpha=0.6) {
type <- match.arg(type)
alternative <- match.arg(alternative)
dmin <- esmin / d_unitconversion
dmax <- esmax / d_unitconversion
values <- pwr.t.test_contourcalc(nmin, nmax, dmin, dmax,
sig.level, type,
alternative)
values$ES = values$ES * d_unitconversion
ylab <- paste0("Hypothetical effect size (", d_units, ")")
xlab <- ifelse(type == "one.sample",
yes="Sample size",
no="Sample size of each group")
create_contour(x = values$n,
y = values$ES,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha)
}
#' Create a Power Table for a t-test comparison
#'
#' Plot a power analysis table of effect size and power for t-test comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # Straightforward example
#' pwr.t.test.table(n=25)
#'
#' # More customised example
#' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
pwr.t.test.table <- function(n,
sig.level=0.05,
type = c("two.sample", "one.sample", "paired"),
alternative = c("two.sided",
"less", "greater"),
d_unitconversion=1,
d_units="Cohen's d",
kable = TRUE,
digits=3, ...) {
type <- match.arg(type)
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
powertable$ES = as.numeric(lapply(powertable$power,
function(x) pwr::pwr.t.test(n=n, power=x,
sig.level = sig.level,
type=type,
alternative=alternative)$d))
powertable$ES <- powertable$ES * d_unitconversion
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
eslevel_extra <- round(c(0, powertable$ES, Inf), digits)
ele_n <- length(eslevel_extra)-1
eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "delta", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
eslevels_text[1] <- paste( "delta", "<", eslevel_extra[2],sep = " ")
eslevels_text[ele_n] <- paste( "delta", ">", eslevel_extra[(ele_n)],sep = " ")
### Pretty table
prettytable <- tibble::tibble(
"True Effect Size" = eslevels_text,
"Power to detect" = power_text,
Description = descriptions_text
)
colnames(prettytable)[1] <- paste0("True Effect Size (", d_units, ")")
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = "Power by effect size", ...)
}
return(prettytable)
}
#' Plot a Power Contour for a correlation analysis
#'
#' Plot a contour of sample size, effect size and power for correlation comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param rmin Minimum Pearson's r for the plot.
#' @param rmax Maximum Pearson's r for the plot.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' pwr.r.test.contour(5, 100, 0.1, 0.7, sig.level=0.05)
pwr.r.test.contour <- function(nmin, nmax, rmin, rmax,
sig.level=0.05,
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
alpha=0.6) {
alternative <- match.arg(alternative)
values <- pwr.r.test_contourcalc(nmin, nmax, rmin, rmax,
sig.level,
alternative)
ylab <- paste0("Hypothetical effect size (Pearson\'s r)")
xlab <- paste0("Sample size")
create_contour(x = values$n,
y = values$ES,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha)
}
#' Create a Power Table for a correlation analysis
#'
#' Plot a power analysis table of effect size and power for Pearson's r tests
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @examples
#' pwr.r.test.table(n=25)
pwr.r.test.table <- function(n,
sig.level=0.05,
alternative = c("two.sided",
"less", "greater"),
kable = TRUE, digits=3, ...) {
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
powertable$ES = as.numeric(lapply(powertable$power,
function(x) pwr::pwr.r.test(n=n, power=x,
sig.level = sig.level,
alternative=alternative)$r))
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
eslevel_extra <- round(c(0, powertable$ES, 1), digits)
ele_n <- length(eslevel_extra)-1
eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "r", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
eslevels_text[1] <- paste( "r", "<", eslevel_extra[2],sep = " ")
eslevels_text[ele_n] <- paste( "r", ">", eslevel_extra[(ele_n)],sep = " ")
### Pretty table
prettytable <- tibble::tibble(
"True Effect Size" = eslevels_text,
"Power to detect" = power_text,
Description = descriptions_text
)
colnames(prettytable)[1] <- "True Effect Size (Pearson's r)"
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = "Power by effect size", ...)
}
return(prettytable)
}
#' Plot a Power Contour for a proportion test
#'
#' Plot a contour of sample size, effect size and power for proportion test
#' comparisons in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param propmin Minimum proportion for the plot.
#' @param propmax Maximum proportion for the plot.
#' @param propcomp Comparison proportion value against which the measured proportion is compared.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The proportion test type. The default is "two.sample", but another option is "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' pwr.prop.test.contour(5, 100, 0.1, 0.9)
pwr.prop.test.contour <- function(nmin, nmax, propmin, propmax, propcomp=0.5,
sig.level=0.05,
type = c("two.sample", "one.sample"),
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
alpha=0.6) {
alternative <- match.arg(alternative)
values <- pwr.prop.test_contourcalc(nmin, nmax, propmin, propmax,
propcomp,
sig.level, type,
alternative)
ylab <- paste0("Hypothetical observed proportion")
xlab <- paste0("Sample size")
create_contour(x = values$n,
y = values$Proportions,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha) +
ggplot2::geom_abline(slope = 0, intercept=propcomp, linetype="dotted")
}
#' Create a Power Table for a proportion test
#'
#' Plot a power analysis table of effect size and power for a proportion test
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param propcomp Comparison proportion value against which the measured proportion is compared.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # Straightforward example
#' pwr.t.test.table(n=25)
#'
#' # More customised example
#' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
pwr.prop.test.table <- function(n, propcomp=0.5,
sig.level=0.05,
type = c("two.sample", "one.sample"),
alternative = c("two.sided",
"less", "greater"),
kable = TRUE,
digits=3, ...) {
type <- match.arg(type)
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
if( type == "two.sample" ) {
powertable$ES <- purrr::map_dbl(powertable$power,
~pwr::pwr.2p.test(n = n, power = .x,
sig.level = sig.level,
alternative = alternative)$h )
} else if( type == "one.sample" ) {
powertable$ES <- purrr::map_dbl(pwr_contours$ES,
~pwr::pwr.p.test(n = n, power = .x,
sig.level = sig.level,
alternative = alternative)$h )
}
powertable_g <- powertable
powertable_l <- powertable
powertable_g$Proportion <- purrr::map_dbl( powertable$ES, ~h2p1(.x, propcomp))
powertable_l$Proportion <- purrr::map_dbl(-powertable$ES, ~h2p1(.x, propcomp))
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
# Greater
eslevel_extra_g <- round(c(0, powertable_g$Proportion, Inf), digits)
ele_n <- length(eslevel_extra_g)-1
eslevels_text_g <- paste(eslevel_extra_g[1:ele_n], "<", "prop", "<", eslevel_extra_g[2:(ele_n+1)],sep = " ")
eslevels_text_g[1] <- paste( "prop", "<", eslevel_extra_g[2],sep = " ")
eslevels_text_g[ele_n] <- paste( "prop", ">", eslevel_extra_g[(ele_n)],sep = " ")
# Less
eslevel_extra_l <- round(c(0, powertable_l$Proportion, Inf), digits)
ele_n <- length(eslevel_extra_l)-1
eslevels_text_l <- paste(eslevel_extra_l[1:ele_n], ">", "prop", ">", eslevel_extra_l[2:(ele_n+1)],sep = " ")
eslevels_text_l[1] <- paste( "prop", ">", eslevel_extra_l[2],sep = " ")
eslevels_text_l[ele_n] <- paste( "prop", "<", eslevel_extra_l[(ele_n)],sep = " ")
### Pretty table
prettytable_g <- tibble::tibble(
"True Proportion" = eslevels_text_g,
"Power to detect" = power_text,
Description = descriptions_text
)
prettytable_l <- tibble::tibble(
"True Proportion" = eslevels_text_l,
"Power to detect" = power_text,
Description = descriptions_text
)
## Combining
if( alternative == "two.sided" ) {
prettytable <- tibble::tibble(
"True Increased Proportion" = eslevels_text_g,
"True Decreased Proportion" = eslevels_text_l,
"Power to detect" = power_text,
Description = descriptions_text)
} else if(alternative == "greater") {
prettytable = prettytable_g
} else if(alternative == "less") {
prettytable = prettytable_l
}
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = paste("Power by proportion, relative to", propcomp), ...)
}
return(prettytable)
}
# #' Plot a Power Contour for a test of equivalence for a t-test
# #'
# #' Plot a contour of sample size, effect size and power for t-test
# #' of equivalence in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param nmin Minimum sample size for the plot.
# #' @param nmax Maximum sample size for the plot.
# #' @param esmin Minimum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
# #' @param esmax Maximum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
# #' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
# #' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the figure are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
# #' @param d_units The units of the difference. Defaults to "Cohen's d".
# #' @param alpha The transparency of the contours, in order to show grid lines below.
# #'
# #' @return A `ggplot2` object.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' # A straightforward example
# #' pwr.eq.t.test.contour(3, 100, 0.1, 1, sig.level=0.05)
# #'
# #' # Custom units
# #' pwr.eq.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
# #' d_unitconversion=5, d_units="mmHg")
# #'
# #' # Adding things
# #' pwr.eq.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
# #' d_unitconversion=5, d_units="mmHg") +
# #' ggplot2::annotate("text", x=25, y=2,
# #' label="We don't want to be here")
# pwr.eq.t.test.contour <- function(nmin, nmax, esmin, esmax,
# sig.level=0.05,
# type = c("two.sample", "one.sample", "paired"),
# pwrbands = c(0.5, 0.8, 0.95),
# d_unitconversion=1,
# d_units="Cohen\'s d",
# alpha=0.6) {
#
#
# type <- match.arg(type)
#
# dmin <- esmin / d_unitconversion
# dmax <- esmax / d_unitconversion
#
# values <- pwr.eq.t.test_contourcalc(nmin, nmax, dmin, dmax,
# sig.level, type)
#
# values$ES = values$ES * d_unitconversion
#
# ylab <- paste0("Hypothetical equivalence bounds (", d_units, ")")
# xlab <- ifelse(type == "one.sample",
# yes="Sample size",
# no="Sample size of each group")
#
# create_contour(x = values$n,
# y = values$ES,
# pwr = values$power,
# pwrcolours = seq(0,1, by=0.1),
# xlab=xlab,
# ylab=ylab,
# pwrbands=pwrbands,
# alpha=alpha)
#
#
# }
#
# #' Create a Power Table for a test of equivalence for a t-test
# #'
# #' Plot a power analysis table of effect size and power for t-test comparisons
# #' in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param n The proposed sample size.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
# #' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
# #' @param d_units The units of the difference. Defaults to "Cohen's d".
# #' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
# #' @param digits The number of digits.
# #' @param ... Extra input arguments for the `knitr::kable()` function
# #'
# #' @return A power table.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' # Straightforward example
# #' pwr.t.test.table(n=25)
# #'
# #' # More customised example
# #' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
# pwr.eq.t.test.table <- function(n,
# sig.level=0.05,
# type = c("two.sample", "one.sample", "paired"),
# d_unitconversion=1,
# d_units="Cohen's d",
# kable = TRUE,
# digits=3, ...) {
#
#
# type <- match.arg(type)
#
# powerlevels = c(0.5, 0.8, 0.95)
#
# powertable <- data.frame(
# power = powerlevels
# )
#
# if(type == "two.sample") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTtwo(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# } else if(type == "paired") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTpaired(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# } else if(type == "one.sample") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTone(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# }
#
# powertable$ES <- powertable$ES * d_unitconversion
#
#
#
# descriptions_text <- c("Likely miss",
# "Good chance of missing",
# "Probably detect",
# "Almost surely detect")
#
# power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#
# #### Text Effect Sizes
#
# eslevel_extra <- round(c(0, powertable$ES, Inf), digits)
# ele_n <- length(eslevel_extra)-1
# eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "delta", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
#
# eslevels_text[1] <- paste( "delta", "<", eslevel_extra[2],sep = " ")
# eslevels_text[ele_n] <- paste( "delta", ">", eslevel_extra[(ele_n)],sep = " ")
#
#
# ### Pretty table
#
# prettytable <- tibble::tibble(
# "True Effect Size" = eslevels_text,
# "Power to detect" = power_text,
# Description = descriptions_text
# )
#
# colnames(prettytable)[1] <- paste0("True Effect Size (", d_units, ")")
#
# if( kable ) {
# prettytable <-
# knitr::kable(prettytable, align = "ccl",
# caption = "Power by effect size", ...)
# }
#
# return(prettytable)
#
# }
#
# #' Plot a Power Contour for a test of equivalence for a correlation
# #'
# #' Plot a contour of sample size, effect size and power for a test of
# #' equivalence for a correlation comparisons in the style of Richard Morey's
# #' jpower JAMOVI module, assuming the true effect is equal to zero.
# #'
# #' @param nmin Minimum sample size for the plot.
# #' @param nmax Maximum sample size for the plot.
# #' @param rmin Minimum Pearson's r for the plot.
# #' @param rmax Maximum Pearson's r for the plot.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param pwrbands Additional black power bands to display on the figure to
# #' emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
# #' @param alpha The transparency of the contours, in order to show grid lines
# #' below.
# #'
# #' @return A `ggplot2` object.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' pwr.r.test.contour(5, 100, 0.1, 0.7, sig.level=0.05)
# pwr.eq.r.test.contour <- function(nmin, nmax, rmin, rmax,
# sig.level=0.05,
# pwrbands = c(0.5, 0.8, 0.95),
# alpha=0.6) {
#
#
# values <- pwr.eq.r.test_contourcalc(nmin, nmax, rmin, rmax,
# sig.level)
#
# ylab <- paste0("Hypothetical equivalence bounds (Pearson\'s r)")
# xlab <- paste0("Sample size")
#
# create_contour(x = values$n,
# y = values$ES,
# pwr = values$power,
# pwrcolours = seq(0,1, by=0.1),
# xlab=xlab,
# ylab=ylab,
# pwrbands=pwrbands,
# alpha=alpha)
#
#
# }
#
#
#
#
# #' Create a Power Table for a for a test of equivalence for a correlation
# #'
# #' Plot a power analysis table of effect size and power for Pearson's r tests
# #' in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param n The proposed sample size.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
# #' @param digits The number of digits.
# #' @param ... Extra input arguments for the `knitr::kable()` function
# #'
# #' @return A power table.
# #' @export
# #'
# #' @examples
# #' pwr.r.test.table(n=25)
# pwr.eq.r.test.table <- function(n,
# sig.level=0.05,
# kable = TRUE, digits=3, ...) {
#
#
#
# powerlevels = c(0.5, 0.8, 0.95)
#
# powertable <- data.frame(
# power = powerlevels
# )
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTr(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
#
#
# descriptions_text <- c("Likely miss",
# "Good chance of missing",
# "Probably detect",
# "Almost surely detect")
#
# power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#
# #### Text Effect Sizes
#
# eslevel_extra <- round(c(0, powertable$ES, 1), digits)
# ele_n <- length(eslevel_extra)-1
# eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "r", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
#
# eslevels_text[1] <- paste( "r", "<", eslevel_extra[2],sep = " ")
# eslevels_text[ele_n] <- paste( "r", ">", eslevel_extra[(ele_n)],sep = " ")
#
#
# ### Pretty table
#
# prettytable <- tibble::tibble(
# "True Effect Size" = eslevels_text,
# "Power to detect" = power_text,
# Description = descriptions_text
# )
#
# colnames(prettytable)[1] <- "True Effect Size (Pearson's r)"
#
# if( kable ) {
# prettytable <-
# knitr::kable(prettytable, align = "ccl",
# caption = "Power by effect size", ...)
# }
#
# return(prettytable)
#
# }
mathesong/pwrcontour documentation built on Jan. 1, 2021, 9:19 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions pwr.prop.test.table pwr.prop.test.contour pwr.r.test.table pwr.r.test.contour pwr.t.test.table pwr.t.test.contour
Documented in pwr.prop.test.contour pwr.prop.test.table pwr.r.test.contour pwr.r.test.table pwr.t.test.contour pwr.t.test.table
#' Plot a Power Contour for a t-test comparison
#'
#' Plot a contour of sample size, effect size and power for t-test comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param esmin Minimum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
#' @param esmax Maximum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the figure are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # A straightforward example
#' pwr.t.test.contour(3, 100, 0.1, 1, sig.level=0.05)
#'
#' # Custom units
#' pwr.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
#' d_unitconversion=5, d_units="mmHg")
#'
#' # Adding things
#' pwr.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
#' d_unitconversion=5, d_units="mmHg") +
#' ggplot2::annotate("text", x=25, y=2,
#' label="We don't want to be here")
pwr.t.test.contour <- function(nmin, nmax, esmin, esmax,
sig.level=0.05,
type = c("two.sample", "one.sample", "paired"),
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
d_unitconversion=1,
d_units="Cohen\'s d",
alpha=0.6) {
type <- match.arg(type)
alternative <- match.arg(alternative)
dmin <- esmin / d_unitconversion
dmax <- esmax / d_unitconversion
values <- pwr.t.test_contourcalc(nmin, nmax, dmin, dmax,
sig.level, type,
alternative)
values$ES = values$ES * d_unitconversion
ylab <- paste0("Hypothetical effect size (", d_units, ")")
xlab <- ifelse(type == "one.sample",
yes="Sample size",
no="Sample size of each group")
create_contour(x = values$n,
y = values$ES,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha)
}
#' Create a Power Table for a t-test comparison
#'
#' Plot a power analysis table of effect size and power for t-test comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # Straightforward example
#' pwr.t.test.table(n=25)
#'
#' # More customised example
#' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
pwr.t.test.table <- function(n,
sig.level=0.05,
type = c("two.sample", "one.sample", "paired"),
alternative = c("two.sided",
"less", "greater"),
d_unitconversion=1,
d_units="Cohen's d",
kable = TRUE,
digits=3, ...) {
type <- match.arg(type)
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
powertable$ES = as.numeric(lapply(powertable$power,
function(x) pwr::pwr.t.test(n=n, power=x,
sig.level = sig.level,
type=type,
alternative=alternative)$d))
powertable$ES <- powertable$ES * d_unitconversion
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
eslevel_extra <- round(c(0, powertable$ES, Inf), digits)
ele_n <- length(eslevel_extra)-1
eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "delta", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
eslevels_text[1] <- paste( "delta", "<", eslevel_extra[2],sep = " ")
eslevels_text[ele_n] <- paste( "delta", ">", eslevel_extra[(ele_n)],sep = " ")
### Pretty table
prettytable <- tibble::tibble(
"True Effect Size" = eslevels_text,
"Power to detect" = power_text,
Description = descriptions_text
)
colnames(prettytable)[1] <- paste0("True Effect Size (", d_units, ")")
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = "Power by effect size", ...)
}
return(prettytable)
}
#' Plot a Power Contour for a correlation analysis
#'
#' Plot a contour of sample size, effect size and power for correlation comparisons
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param rmin Minimum Pearson's r for the plot.
#' @param rmax Maximum Pearson's r for the plot.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' pwr.r.test.contour(5, 100, 0.1, 0.7, sig.level=0.05)
pwr.r.test.contour <- function(nmin, nmax, rmin, rmax,
sig.level=0.05,
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
alpha=0.6) {
alternative <- match.arg(alternative)
values <- pwr.r.test_contourcalc(nmin, nmax, rmin, rmax,
sig.level,
alternative)
ylab <- paste0("Hypothetical effect size (Pearson\'s r)")
xlab <- paste0("Sample size")
create_contour(x = values$n,
y = values$ES,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha)
}
#' Create a Power Table for a correlation analysis
#'
#' Plot a power analysis table of effect size and power for Pearson's r tests
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @examples
#' pwr.r.test.table(n=25)
pwr.r.test.table <- function(n,
sig.level=0.05,
alternative = c("two.sided",
"less", "greater"),
kable = TRUE, digits=3, ...) {
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
powertable$ES = as.numeric(lapply(powertable$power,
function(x) pwr::pwr.r.test(n=n, power=x,
sig.level = sig.level,
alternative=alternative)$r))
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
eslevel_extra <- round(c(0, powertable$ES, 1), digits)
ele_n <- length(eslevel_extra)-1
eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "r", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
eslevels_text[1] <- paste( "r", "<", eslevel_extra[2],sep = " ")
eslevels_text[ele_n] <- paste( "r", ">", eslevel_extra[(ele_n)],sep = " ")
### Pretty table
prettytable <- tibble::tibble(
"True Effect Size" = eslevels_text,
"Power to detect" = power_text,
Description = descriptions_text
)
colnames(prettytable)[1] <- "True Effect Size (Pearson's r)"
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = "Power by effect size", ...)
}
return(prettytable)
}
#' Plot a Power Contour for a proportion test
#'
#' Plot a contour of sample size, effect size and power for proportion test
#' comparisons in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param nmin Minimum sample size for the plot.
#' @param nmax Maximum sample size for the plot.
#' @param propmin Minimum proportion for the plot.
#' @param propmax Maximum proportion for the plot.
#' @param propcomp Comparison proportion value against which the measured proportion is compared.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The proportion test type. The default is "two.sample", but another option is "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
#' @param alpha The transparency of the contours, in order to show grid lines below.
#'
#' @return A `ggplot2` object.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' pwr.prop.test.contour(5, 100, 0.1, 0.9)
pwr.prop.test.contour <- function(nmin, nmax, propmin, propmax, propcomp=0.5,
sig.level=0.05,
type = c("two.sample", "one.sample"),
alternative = c("two.sided",
"less", "greater"),
pwrbands = c(0.5, 0.8, 0.95),
alpha=0.6) {
alternative <- match.arg(alternative)
values <- pwr.prop.test_contourcalc(nmin, nmax, propmin, propmax,
propcomp,
sig.level, type,
alternative)
ylab <- paste0("Hypothetical observed proportion")
xlab <- paste0("Sample size")
create_contour(x = values$n,
y = values$Proportions,
pwr = values$power,
pwrcolours = seq(0,1, by=0.1),
xlab=xlab,
ylab=ylab,
pwrbands=pwrbands,
alpha=alpha) +
ggplot2::geom_abline(slope = 0, intercept=propcomp, linetype="dotted")
}
#' Create a Power Table for a proportion test
#'
#' Plot a power analysis table of effect size and power for a proportion test
#' in the style of Richard Morey's jpower JAMOVI module.
#'
#' @param n The proposed sample size.
#' @param propcomp Comparison proportion value against which the measured proportion is compared.
#' @param sig.level Significance level, or alpha. Defaults to 0.05.
#' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
#' @param alternative The alternative hypothesis. The default is "two.sided", but other options are "less" or "greater".
#' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
#' @param d_units The units of the difference. Defaults to "Cohen's d".
#' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
#' @param digits The number of digits.
#' @param ... Extra input arguments for the `knitr::kable()` function
#'
#' @return A power table.
#' @export
#'
#' @author Granville J Matheson, \email{mathesong@@gmail.com}
#'
#' @examples
#' # Straightforward example
#' pwr.t.test.table(n=25)
#'
#' # More customised example
#' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
pwr.prop.test.table <- function(n, propcomp=0.5,
sig.level=0.05,
type = c("two.sample", "one.sample"),
alternative = c("two.sided",
"less", "greater"),
kable = TRUE,
digits=3, ...) {
type <- match.arg(type)
alternative <- match.arg(alternative)
powerlevels = c(0.5, 0.8, 0.95)
powertable <- data.frame(
power = powerlevels
)
if( type == "two.sample" ) {
powertable$ES <- purrr::map_dbl(powertable$power,
~pwr::pwr.2p.test(n = n, power = .x,
sig.level = sig.level,
alternative = alternative)$h )
} else if( type == "one.sample" ) {
powertable$ES <- purrr::map_dbl(pwr_contours$ES,
~pwr::pwr.p.test(n = n, power = .x,
sig.level = sig.level,
alternative = alternative)$h )
}
powertable_g <- powertable
powertable_l <- powertable
powertable_g$Proportion <- purrr::map_dbl( powertable$ES, ~h2p1(.x, propcomp))
powertable_l$Proportion <- purrr::map_dbl(-powertable$ES, ~h2p1(.x, propcomp))
descriptions_text <- c("Likely miss",
"Good chance of missing",
"Probably detect",
"Almost surely detect")
power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#### Text Effect Sizes
# Greater
eslevel_extra_g <- round(c(0, powertable_g$Proportion, Inf), digits)
ele_n <- length(eslevel_extra_g)-1
eslevels_text_g <- paste(eslevel_extra_g[1:ele_n], "<", "prop", "<", eslevel_extra_g[2:(ele_n+1)],sep = " ")
eslevels_text_g[1] <- paste( "prop", "<", eslevel_extra_g[2],sep = " ")
eslevels_text_g[ele_n] <- paste( "prop", ">", eslevel_extra_g[(ele_n)],sep = " ")
# Less
eslevel_extra_l <- round(c(0, powertable_l$Proportion, Inf), digits)
ele_n <- length(eslevel_extra_l)-1
eslevels_text_l <- paste(eslevel_extra_l[1:ele_n], ">", "prop", ">", eslevel_extra_l[2:(ele_n+1)],sep = " ")
eslevels_text_l[1] <- paste( "prop", ">", eslevel_extra_l[2],sep = " ")
eslevels_text_l[ele_n] <- paste( "prop", "<", eslevel_extra_l[(ele_n)],sep = " ")
### Pretty table
prettytable_g <- tibble::tibble(
"True Proportion" = eslevels_text_g,
"Power to detect" = power_text,
Description = descriptions_text
)
prettytable_l <- tibble::tibble(
"True Proportion" = eslevels_text_l,
"Power to detect" = power_text,
Description = descriptions_text
)
## Combining
if( alternative == "two.sided" ) {
prettytable <- tibble::tibble(
"True Increased Proportion" = eslevels_text_g,
"True Decreased Proportion" = eslevels_text_l,
"Power to detect" = power_text,
Description = descriptions_text)
} else if(alternative == "greater") {
prettytable = prettytable_g
} else if(alternative == "less") {
prettytable = prettytable_l
}
if( kable ) {
prettytable <-
knitr::kable(prettytable, align = "ccl",
caption = paste("Power by proportion, relative to", propcomp), ...)
}
return(prettytable)
}
# #' Plot a Power Contour for a test of equivalence for a t-test
# #'
# #' Plot a contour of sample size, effect size and power for t-test
# #' of equivalence in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param nmin Minimum sample size for the plot.
# #' @param nmax Maximum sample size for the plot.
# #' @param esmin Minimum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
# #' @param esmax Maximum effect size for the plot. Defaults to Cohen's d, but can be percentages or units of measurement if desired using the `d_unitconversion` parameter.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
# #' @param pwrbands Additional black power bands to display on the figure to emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
# #' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the figure are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
# #' @param d_units The units of the difference. Defaults to "Cohen's d".
# #' @param alpha The transparency of the contours, in order to show grid lines below.
# #'
# #' @return A `ggplot2` object.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' # A straightforward example
# #' pwr.eq.t.test.contour(3, 100, 0.1, 1, sig.level=0.05)
# #'
# #' # Custom units
# #' pwr.eq.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
# #' d_unitconversion=5, d_units="mmHg")
# #'
# #' # Adding things
# #' pwr.eq.t.test.contour(3, 100, 0.5, 5, sig.level=0.05,
# #' d_unitconversion=5, d_units="mmHg") +
# #' ggplot2::annotate("text", x=25, y=2,
# #' label="We don't want to be here")
# pwr.eq.t.test.contour <- function(nmin, nmax, esmin, esmax,
# sig.level=0.05,
# type = c("two.sample", "one.sample", "paired"),
# pwrbands = c(0.5, 0.8, 0.95),
# d_unitconversion=1,
# d_units="Cohen\'s d",
# alpha=0.6) {
#
#
# type <- match.arg(type)
#
# dmin <- esmin / d_unitconversion
# dmax <- esmax / d_unitconversion
#
# values <- pwr.eq.t.test_contourcalc(nmin, nmax, dmin, dmax,
# sig.level, type)
#
# values$ES = values$ES * d_unitconversion
#
# ylab <- paste0("Hypothetical equivalence bounds (", d_units, ")")
# xlab <- ifelse(type == "one.sample",
# yes="Sample size",
# no="Sample size of each group")
#
# create_contour(x = values$n,
# y = values$ES,
# pwr = values$power,
# pwrcolours = seq(0,1, by=0.1),
# xlab=xlab,
# ylab=ylab,
# pwrbands=pwrbands,
# alpha=alpha)
#
#
# }
#
# #' Create a Power Table for a test of equivalence for a t-test
# #'
# #' Plot a power analysis table of effect size and power for t-test comparisons
# #' in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param n The proposed sample size.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param type The t-test type. The default is "two.sample", but other options are "paired" or "one.sample".
# #' @param d_unitconversion Unit conversion of Cohen's d. If specified as 5, for example, 5 units of change in the units of the table are considered equivalent to Cohen's d = 1. This conversion should be calculated in advance. Defaults to 1 for Cohen's d.
# #' @param d_units The units of the difference. Defaults to "Cohen's d".
# #' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
# #' @param digits The number of digits.
# #' @param ... Extra input arguments for the `knitr::kable()` function
# #'
# #' @return A power table.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' # Straightforward example
# #' pwr.t.test.table(n=25)
# #'
# #' # More customised example
# #' pwr.t.test.table(n=25, d_unitconversion=20, d_units="%", digits=1)
# pwr.eq.t.test.table <- function(n,
# sig.level=0.05,
# type = c("two.sample", "one.sample", "paired"),
# d_unitconversion=1,
# d_units="Cohen's d",
# kable = TRUE,
# digits=3, ...) {
#
#
# type <- match.arg(type)
#
# powerlevels = c(0.5, 0.8, 0.95)
#
# powertable <- data.frame(
# power = powerlevels
# )
#
# if(type == "two.sample") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTtwo(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# } else if(type == "paired") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTpaired(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# } else if(type == "one.sample") {
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTone(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
# }
#
# powertable$ES <- powertable$ES * d_unitconversion
#
#
#
# descriptions_text <- c("Likely miss",
# "Good chance of missing",
# "Probably detect",
# "Almost surely detect")
#
# power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#
# #### Text Effect Sizes
#
# eslevel_extra <- round(c(0, powertable$ES, Inf), digits)
# ele_n <- length(eslevel_extra)-1
# eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "delta", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
#
# eslevels_text[1] <- paste( "delta", "<", eslevel_extra[2],sep = " ")
# eslevels_text[ele_n] <- paste( "delta", ">", eslevel_extra[(ele_n)],sep = " ")
#
#
# ### Pretty table
#
# prettytable <- tibble::tibble(
# "True Effect Size" = eslevels_text,
# "Power to detect" = power_text,
# Description = descriptions_text
# )
#
# colnames(prettytable)[1] <- paste0("True Effect Size (", d_units, ")")
#
# if( kable ) {
# prettytable <-
# knitr::kable(prettytable, align = "ccl",
# caption = "Power by effect size", ...)
# }
#
# return(prettytable)
#
# }
#
# #' Plot a Power Contour for a test of equivalence for a correlation
# #'
# #' Plot a contour of sample size, effect size and power for a test of
# #' equivalence for a correlation comparisons in the style of Richard Morey's
# #' jpower JAMOVI module, assuming the true effect is equal to zero.
# #'
# #' @param nmin Minimum sample size for the plot.
# #' @param nmax Maximum sample size for the plot.
# #' @param rmin Minimum Pearson's r for the plot.
# #' @param rmax Maximum Pearson's r for the plot.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param pwrbands Additional black power bands to display on the figure to
# #' emphasise certain power levels. Defaults to c(0.5, 0.8, 0.95).
# #' @param alpha The transparency of the contours, in order to show grid lines
# #' below.
# #'
# #' @return A `ggplot2` object.
# #' @export
# #'
# #' @author Granville J Matheson, \email{mathesong@@gmail.com}
# #'
# #' @examples
# #' pwr.r.test.contour(5, 100, 0.1, 0.7, sig.level=0.05)
# pwr.eq.r.test.contour <- function(nmin, nmax, rmin, rmax,
# sig.level=0.05,
# pwrbands = c(0.5, 0.8, 0.95),
# alpha=0.6) {
#
#
# values <- pwr.eq.r.test_contourcalc(nmin, nmax, rmin, rmax,
# sig.level)
#
# ylab <- paste0("Hypothetical equivalence bounds (Pearson\'s r)")
# xlab <- paste0("Sample size")
#
# create_contour(x = values$n,
# y = values$ES,
# pwr = values$power,
# pwrcolours = seq(0,1, by=0.1),
# xlab=xlab,
# ylab=ylab,
# pwrbands=pwrbands,
# alpha=alpha)
#
#
# }
#
#
#
#
# #' Create a Power Table for a for a test of equivalence for a correlation
# #'
# #' Plot a power analysis table of effect size and power for Pearson's r tests
# #' in the style of Richard Morey's jpower JAMOVI module,
# #' assuming the true effect is equal to zero.
# #'
# #' @param n The proposed sample size.
# #' @param sig.level Significance level, or alpha. Defaults to 0.05.
# #' @param kable Should the table be presented as a `knitr::kable()` figure? Defaults to `TRUE`. Otherwise a `tibble` is returned.
# #' @param digits The number of digits.
# #' @param ... Extra input arguments for the `knitr::kable()` function
# #'
# #' @return A power table.
# #' @export
# #'
# #' @examples
# #' pwr.r.test.table(n=25)
# pwr.eq.r.test.table <- function(n,
# sig.level=0.05,
# kable = TRUE, digits=3, ...) {
#
#
#
# powerlevels = c(0.5, 0.8, 0.95)
#
# powertable <- data.frame(
# power = powerlevels
# )
#
# powertable$ES <- purrr::map_dbl(powertable$power,
# ~quiet(TOSTER::powerTOSTr(N = n,
# statistical_power = .x,
# alpha = sig.level)[1] ))
#
#
# descriptions_text <- c("Likely miss",
# "Good chance of missing",
# "Probably detect",
# "Almost surely detect")
#
# power_text <- c("< 50%", "50% - 80%", "80% - 95%", "> 95%")
#
# #### Text Effect Sizes
#
# eslevel_extra <- round(c(0, powertable$ES, 1), digits)
# ele_n <- length(eslevel_extra)-1
# eslevels_text <- paste(eslevel_extra[1:ele_n], "<", "r", "<", eslevel_extra[2:(ele_n+1)],sep = " ")
#
# eslevels_text[1] <- paste( "r", "<", eslevel_extra[2],sep = " ")
# eslevels_text[ele_n] <- paste( "r", ">", eslevel_extra[(ele_n)],sep = " ")
#
#
# ### Pretty table
#
# prettytable <- tibble::tibble(
# "True Effect Size" = eslevels_text,
# "Power to detect" = power_text,
# Description = descriptions_text
# )
#
# colnames(prettytable)[1] <- "True Effect Size (Pearson's r)"
#
# if( kable ) {
# prettytable <-
# knitr::kable(prettytable, align = "ccl",
# caption = "Power by effect size", ...)
# }
#
# return(prettytable)
#
# }
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.